지극히 개인적인 일반적인 구조의 CNN 전용 template
- Classification, Segmentation 에 적합한(?) template
-
data
- Dataset Name
- train
- validation
- test
- Dataset Name
-
log
- Experiment name
- date
- ...
- date
- Experiment name
-
model
- loss.py
- metric.py
- model.py
-
trainer
- dataloader.py
- evaluation.py
- training.py
-
utils
- ckpt.py
- log.py
- Using Package
- torch==1.7.1
- torchvision==0.8.2
- torchaudio==0.7.2
# Create environment
conda create -y -n torch python=3.7
conda activate torch
# CPU mode
pip install --no-cache-dir torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
pip install -r requirements.txt
# GPU mode
pip install --no-cache-dir torch torchvision torchaudio
pip install -r requirements.txt
- Download example dataset
- Unzip dataset in
data
directory - Edit
config.yaml
- Execute command
python train.py --config config.yaml
...ing
- Download example dataset
- Multi GPU
- Inference code
- Weights & Biases 연결?
- Segmentation